Project Summary: Over the last 30 years, spine pain has been the most common reason for Medicare enrollment based on disability, and Medicare enrollees ? whether enrolled by virtue of disability or age ? have increasingly used spinal manipulation therapy to treat such pain. The American Heart and American Stroke Associations (AHA/ASA) recommended that patients be informed of the statistical association between cervical spinal manipulation (CSM) and cervical artery dissection (CAD) but could not articulate the magnitude of that risk. The literature on which the AHA/ASA based their recommendations used older data from outside the US and flawed methods (inaccurate identification of the primary outcome of interest (CAD), reliance on patient recall to identify CSM, using controls that were considerably healthier than cases, and limiting CSM to that provided by doctors of chiropractic, although other practitioner types also provide CSM). To inform Medicare enrollees, providers, and policymakers about the risk of CAD following CSM, we will analyze Medicare services data from 2004-2015 for beneficiaries concurrently enrolled in Medicare Parts A and B. To address the limitations of prior studies, we will use diagnostic codes specific to CAD to accurately identify cases, match cases to proper controls, identify common comorbidities of CAD, assess outcomes (including mortality and changes in costs of care), and use sophisticated statistical methods to determine whether there is an association between CSM and CAD, by provider type. If we find such a relationship, we will calculate the risk of CAD following a visit for CSM. To avoid the selection bias that occurred in prior studies, we will use propensity scoring both to adjust for the likelihood of obtaining CSM (studies have shown that patients who use CSM are younger, healthier, and wealthier than those who do not) and to match controls on comorbidities found to be associated with CAD. Those adjustments ? in concert with an instrumental variable approach ? should unmask relationships that might have gone undetected in previous studies due to a combination of exposure, selection, and recall bias. Our pilot work indicates that there will be well over 9,000 cases of CAD in the Medicare population during the time period of our analysis, dwarfing the size of all previous studies of CAD. Policymakers, providers, and patients will be able to use our findings about the risks of CSM compared to other commonly used treatments for spine pain to make informed financing and treatment decisions.